futex: Sanitize exit state handling
authorThomas Gleixner <tglx@linutronix.de>
Wed, 6 Nov 2019 21:55:42 +0000 (22:55 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 29 Nov 2019 09:10:12 +0000 (10:10 +0100)
commit 4a8e991b91aca9e20705d434677ac013974e0e30 upstream.

Instead of having a smp_mb() and an empty lock/unlock of task::pi_lock move
the state setting into to the lock section.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20191106224556.645603214@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernel/futex.c

index 777a226..b2e2d0f 100644 (file)
@@ -3703,16 +3703,19 @@ void futex_exit_recursive(struct task_struct *tsk)
 
 void futex_exit_release(struct task_struct *tsk)
 {
-       tsk->futex_state = FUTEX_STATE_EXITING;
-       /*
-        * Ensure that all new tsk->pi_lock acquisitions must observe
-        * FUTEX_STATE_EXITING. Serializes against attach_to_pi_owner().
-        */
-       smp_mb();
        /*
-        * Ensure that we must observe the pi_state in exit_pi_state_list().
+        * Switch the state to FUTEX_STATE_EXITING under tsk->pi_lock.
+        *
+        * This ensures that all subsequent checks of tsk->futex_state in
+        * attach_to_pi_owner() must observe FUTEX_STATE_EXITING with
+        * tsk->pi_lock held.
+        *
+        * It guarantees also that a pi_state which was queued right before
+        * the state change under tsk->pi_lock by a concurrent waiter must
+        * be observed in exit_pi_state_list().
         */
        raw_spin_lock_irq(&tsk->pi_lock);
+       tsk->futex_state = FUTEX_STATE_EXITING;
        raw_spin_unlock_irq(&tsk->pi_lock);
 
        futex_exec_release(tsk);